Benettin Giancarlo
Università di Padua, Dipartimento di Matematica Pura e Applicata, Via G. Belzoni 7, 35131 Padua, Italy.
Chaos. 2005 Mar;15(1):15108. doi: 10.1063/1.1854278.
The FPU problem, i.e., the problem of energy equipartition among normal modes in a weakly nonlinear lattice, is here studied in dimension two, more precisely in a model with triangular cell and nearest-neighbors Lennard-Jones interaction. The number n of degrees of freedom ranges from 182 to 6338. Energy is initially equidistributed among a small number n(0) of low frequency modes, with n(0) proportional to n. We study numerically the time evolution of the so-called spectral entropy and the related "effective number" n(eff) of degrees of freedom involved in the dynamics; in this (rather typical) way we can estimate, for each n and each specific energy (energy per degree of freedom) epsilon, the time scale T(n)(epsilon) for energy equipartition. Numerical results indicate that in the thermodynamic limit the equipartition times are short: more precisely, for large n at fixed epsilon we find a limit curve T(infinity)(epsilon), and T(infinity) grows only as epsilon(-1) for small epsilon. Larger equipartition times are obtained by lowering epsilon, at fixed n, below a crossover value epsilon(c)(n). However, epsilon(c) appears to vanish by increasing n (faster than 1n), and the total energy E=nepsilon, rather than epsilon, appears to be the relevant variable when n is large and epsilon<epsilon(c). In conclusion, it seems that in the thermodynamic limit, for this model and this kind of initial conditions, the FPU phenomenon, namely the lack of energy equipartition in physically reasonable times, practically disappears.
本文研究二维空间中的FPU问题,即弱非线性晶格中正常模式间的能量均分问题,更确切地说是研究具有三角形晶格单元和最近邻 Lennard-Jones 相互作用的模型。自由度的数量 n 范围从182到6338。能量最初在少量 n(0) 个低频模式间均匀分布,其中 n(0) 与 n 成正比。我们通过数值方法研究了所谓的谱熵的时间演化以及动力学中涉及的相关“有效自由度”数量 n(eff);以这种(相当典型的)方式,我们可以针对每个 n 和每个特定能量(每自由度能量)ε,估计能量均分的时间尺度 T(n)(ε)。数值结果表明,在热力学极限下,均分时间很短:更确切地说,对于固定 ε 下的大 n,我们得到一条极限曲线 T(∞)(ε),并且对于小 ε,T(∞) 仅随 ε^(-1) 增长。在固定 n 时,将 ε 降低到交叉值 ε(c)(n) 以下会得到更长的均分时间。然而,ε(c) 似乎随着 n 的增加而消失(比 1/n 更快),并且当 n 很大且 ε < ε(c) 时,总能量 E = nε 而非 ε 似乎是相关变量。总之,对于该模型和这种初始条件,在热力学极限下,FPU 现象,即在物理上合理的时间内缺乏能量均分,实际上似乎消失了。